Speaking during an interview on CNBC’s “Squawk on the Street” segment earlier this week, CEO of cybersecurity giant Palo Alto Networks Nikesh Arora implored the tech industry to lower the cost of AI.
During the segment, the chief executive argued that the cost to use large language models (LLMs) has to drop by 20 percent by 2027 — and 90 percent by 2028 — for the tech to be useful to enterprises.
“We need to see the pricing for AI come down,” Arora said.
Don’t forget anyone with 3 functioning brain cells knows you don’t replace workers with AI, and even using it as a productivity tool can cause workers to offload their jobs onto their coworkers with AI by outsourcing comprehension and cognition with workslop.
The “strong” tasks for AI are so limited in scope I don’t think it’s as world shaking as people think it is. Until they figure out the hallucination thing its just an intern who’s a pathological liar. In fact I think it’s worse since the intern at least knows what the truth is when they try to hide it with a lie. LLMs have no concept of truth or falsehood they just create a confident most-likely answer with no idea of what’s true or false.
They can’t “figure out” the hallucination thing because literally everything that comes out of an LLMbecile is hallucination. It’s just that sometimes (and not as many times as people seem to think!) the hallucination happens to match something in reality.
So why do so many people believe that LLMs “know” things and aren’t constantly hallucinating? We’re getting our cognition hacked.
See, we have no way to directly detect intelligence in other beings. What we have are proxies that have, over millions of years of evolution, proved “good enough” a substitute for the actual measurement to be usable as that measurement. And one of the biggest pieces of proxied intellectual detection we have is … fluency. (You know all those bigoted “Polish jokes” or, from my childhood, “Ukrainian jokes”, or such over the years? Yep. Entire ethnic groups were painted with a broad brush as being “stupid” because they couldn’t speak English well. Thank you proxied intellect measurement!)
LLMs are something unprecedented in the entire history of life on this planet: fluency paired with zero intellect. But we’ve evolved to use fluency as a very strong proxy measure of intellect. LLMs, with their confident wording and their almost supernatural fluency fool us into thinking they’re, well, thinking. And because we assume that they’re smart, we overlook their errors—often helping them along and prompting them down the right path when they hallucinate incorrectly, but close to correct—and when we give them the answer and they state it with fluency and confidence, we’re amazed at how smart they are!
But the truth is that they cannot ever figure out the hallucination thing. It would take a completely different technology from LLMs to solve that problem. LLMs are a dead end for genuine machine intellect.
oh yea only if the hallucination sources from veribel fact sources, like legitimate research papers, official medical/govt,etc sites. but most of the time its REDDIT, blogs, or other forums.
No. You are making a mistake. Here are two sentences.
The orange tree is a tall plant that produces sweet fruits. The giraffe is a tall animal that eats sweet fruits.
These are both true statements. But using statistical generation I could easily, from these, generate “The orange tree is a tall animal that eats sweet fruits.”
EVERYTHING that comes from an LLM, no matter how it is trained, is hallucinatory. NO EXCEPTIONS. The more often that the statistics match reality, the more likely it is that it will match reality, but even if 100% of the input is factual there is always a path in the stochastic generation that will yield (highly fluent) garbage.
Also, remember that Piltdown Man was written about as a real thing in the scientific journal Nature…
I think that it’s more simple to just say that LLMs are highy advanced electronic parrots - they put sentences together out of sentences that they’ve been exposed to before, the more often they’ve been exposed to them the more likely they’ll produced it.
LLMs and Parrots are both just copying stuff without understanding it, it’s just that the LLM has a very deep multi-level mathematical engine to find correlations between words and even between word assemblies of various sizes, all in much larger windows of text (literally within tens of tousands and even hundreds of thousands of words) whilst the Parrot’s “correlation detection” isn’t mathematically strict or all that deep or wide (about enough to output “Polly wants a cracker” if they hear the word “cracker”).
Curiously, of both the Parrot is the one which actually has some cognition (i.e. they think), whilst the LLM is a purelly statistical engine.
IMHO the whole use of “hallucination” just muddles the waters because when we hear the word “hallucination” we think the human version of it (seeing that which isn’t there) which makes us think of human cognition because it’s generally a cognitive problem, whilst the LLM hallucination is something very different - producing a seemingly truthful piece of text which is in fact not truthful.
LLMs don’t hallucinate in the traditional sense of the word because as unthinking statistical text analysis and production engines they have no cognition to malfunction in that way.
So in a way by using the word “hallucination” for LLMs we are indirectly supporting in people’s minds the false idea that LLMs have cognition.
Highly fluent crap is what I get a lot from chatgpt.
Omg I have seen this so much in my last job. Everyone else blindly trusting the LLM while I was calling it a fucking moron because it was wrong most of the time and when it was right it still couldn’t solve the actual problem. Even common issues it frequently fucked up with.
The really insidious way to abuse it is to summarise a report youre supposed to read into an email you send to your colleague. Now they are the ones that have to do all of the reading to make sure it’s right. Same with making and interpreting emails. The thorough person who actually verifies everything is doing outsourced comprehension and cognition from a colleague who didn’t do that to appear more productive.
Also the intern can learn or be reprimanded. With the LLM it’ll just go out of context and you have to keep repeating yourself.
But it was a great way for tech to justify laying off staff as a cover for killing product lines not making money. Then non-tech companies bought the bullshit.
Famous non tech companies such as Palo Alto